EMRFD Message Archive 13140

Message Date From Subject
13140 2016-09-09 07:11:20 jwolczanski Gilbert Cell output
I've prototyped the discrete component version of the NE-602, as described in EMRFD (figure 5.28) and I wonder about the output circuitry (mine is a 20 meter DC receiver).  Unlike a diode ring mixer, I don't see any conversation about the output termination needing to terminate DC-to-daylight.

I see low-pass resistor/capacitor combinations and even, in one case, a series inductor/capacitor filter resonating at the desired audio frequency....600Hz in my case.

Comments/suggestions?

Jerry
KI4IO
Warrenton, VA


13141 2016-09-09 07:22:27 Chris Trask Re: Gilbert Cell output
The output terminals of the Gilbert cell type mixers are open collectors, so the output termination is somewhat arbitrary. The equivalent shunt capacitance and series resistance needs to be taken into account in order to design for the desired bandwidth.

For an audio output, you could use an audio transformer with a high turns ratio to provide a lot of voltage gain. It would need to have a centre tap on the mixer side so as to get the best possible dynamic range.

>
>I've prototyped the discrete component version of the NE-602, as described
>in EMRFD (figure 5.28) and I wonder about the output circuitry (mine is a 20
>meter DC receiver). Unlike a diode ring mixer, I don't see any conversation
>about the output termination needing to terminate DC-to-daylight.
>
>I see low-pass resistor/capacitor combinations and even, in one case, a
>series inductor/capacitor filter resonating at the desired audio
>frequency....600Hz in my case.
>
>Comments/suggestions?
>


Chris Trask
N7ZWY / WDX3HLB
Senior Member IEEE
http://www.home.earthlink.net/~christrask/
13142 2016-09-09 07:26:03 Bill Carver Re: Gilbert Cell output
Unlike the diode mixer, where RF, LO and IF signal all run in the same wires and diodes, each signal has their own port in the NE602 so it's a totally different animal and isn't particularly termination sensitive. That said, you wouldn't want a tuned circuit resonant at the LO frequency at the IF port because there isn't PERFECT isolation of the innards from something that drastic on an output pin, but otherwise it is pretty termination insensitive as long as the output signal doesn't produce so much signal the output saturates. In the case of a resonant termination, just put at least some capacitance from the output pin to ground so LO and RF signals see an RF bypass to ground.

The output pin has a nominal 1K resistance, and that will be in series with your series LC audio filter and reduce the Q. May/may not be a concern.

Bill - W7AAZ


13143 2016-09-09 13:29:24 kb1gmx Re: Gilbert Cell output
The 602/612/1496 and the equivalent circuit are termination insensitive however if you want good gain and performance  the following are suggested.

Input in most all of those its about 3kohm balanged and mor edue to resistors present that the transistors.
Feeding the device balanced has better gain and lower noise.

LO port, its insensitive as its actually isolated from the other ports (the EMRFD version is less so).
It is level  sensitive for best gain or lowest spur production.

Output ports best gain is when this pair runs a balanced transformer or a floating one.  For the 602/612
the impedance is about 3K port to port or 1.5K to ground (unbalanced).  The 1496 and the discrete version
there are no resistors present (or at the users option)so the load can be a ceneter tapped transformer balanced or a single ended output.  In all cases the gain is higher with a balanced output.

The unique feature of all the multipliers (Gilbert cells) is that eachout put can be used independently at 
the sum and the difference freuency at the same time.  This is widely used in simple or compact transceivers.

The greatest difference is that a DBM with amplifiers offers about the same performane at higher power consumption.  The 612 and friends are about 15-17db of gain, a bare DBM is 5-7db of loss.  

There is a AD device the AD831 that does 50 ohm ports and far higher intercept points but the price is 
a lot of current(power) to feed it.   

Allison 
13144 2016-09-10 08:08:30 swift_glen Re: Gilbert Cell output
Jerry,
About your termination-insensitive question. Diode ring mixers couple ports rather tightly, as evidenced by folks swapping ports, and still having a good working mixer. So a mis-match at one port is greatly noticed at other ports.

With your Gilbert cell of EMRFD fig.5.28, your audio output port couples back to L.O. port through Q1,Q2,Q3,Q4 base-collectors. This coupling is mostly through transistor base-to-collector capacitance, which is quite small. L.O. port couples to R.F.input port similarly, through base-to-collector capacitance of Q5,Q6. Again, small coupling. Another factor influencing coupling is circuit balance: Q1,Q2,Q5 vs. Q3,Q4,Q6. With little port-to-port coupling, you have much freedom to choose each port impedance without disturbing others.

Others have advised a balanced audio transformer (to replace the FT37-43 transformer of fig.5.28) as a preferred output termination for the audio output of this mixer. Since there is SOME port-to-port coupling, they've cautioned that your audio port should also include a termination for RF & LO. This is easy in a DC receiver - simply bypass both audio outputs (collectors of Q1, Q2, Q3, Q4) to ground with small capacitances, forming a low-pass filter. Since fig.5.28 shows a wide-band termination, these capacitors are not used.

A simpler output duplicates NE602 methods of fig.5.26: equal-value collector resistors to +supply. Again, add two small bypass caps. The following audio amp should ideally be balanced, like a LM386, or a properly-done differential op-amp. The audio transformer is more versatile, has potentially more gain, but is difficult and expensive to source.
13177 2016-09-23 17:51:09 jwolczanski Re: Gilbert Cell output
Thanks all for your advice regarding my build of a discrete component NE602 (Gilbert Cell) as shown in EMRFD (figure 5.28).

She's up and running.  In my junk box I found a 2k:10K center-tapped transformer and am using the center-tapped 2K side on the output of the "NE602".  I'm following that with the pre-amp from the R1/R2 series and a 301AH op-amp with about 40dB of gain...probably 60dB all total.  I'm getting surprising headphone volume on CW/RTTY signals on 20 meters.

I also employed by-pass caps (.01) on the Collectors.

Question:  Should I be using 1500 ohm resistors vice the 36 ohm resistors as shown in the article (?), since I'm using a audio transformer vice the ferrite coil - I wonder if that would be the case.

I had to make a circuit board - my perf board prototype was a great BCB "antenna" - har

Jerry
KI4IO
Warrenton, VA
13178 2016-09-23 18:51:49 Chris Trask Re: Gilbert Cell output
Glad to hear that you got it working. I used an NE602 and a number of other parts to make a minimum component 10MHz WWV receiver a few years ago.

>
>Thanks all for your advice regarding my build of a discrete component
>NE602 (Gilbert Cell) as shown in EMRFD (figure 5.28).
>
>She's up and running. In my junk box I found a 2k:10K center-tapped
>transformer and am using the center-tapped 2K side on the output of the
>"NE602". I'm following that with the pre-amp from the R1/R2 series and
>a 301AH op-amp with about 40dB of gain...probably 60dB all total. I'm
>getting surprising headphone volume on CW/RTTY signals on 20 meters.
>
>I also employed by-pass caps (.01) on the Collectors.
>
>Question: Should I be using 1500 ohm resistors vice the 36 ohm resistors
>as shown in the article (?), since I'm using a audio transformer vice the
>ferrite coil - I wonder if that would be the case.
>
>I had to make a circuit board - my perf board prototype was a great BCB
>"antenna" - har
>


Chris Trask
N7ZWY / WDX3HLB
Senior Member IEEE
http://www.home.earthlink.net/~christrask/
13180 2016-09-25 10:01:18 kb1gmx Re: Gilbert Cell output
If your audio transformer is wired like the RF one in the example then leave the 36 ohms resistors.

I rarely use perf board at RF.  I dead bug/ugly/manhatten on a random piece of copper clad.

Allison
13181 2016-09-25 10:47:10 jwolczanski Re: Gilbert Cell output
Thanks Allison

I used a 2K (center-tap) to 10K transformer from my junk box initially on the output of the "NE602".  This morning I swapped to a [normal] 1000 ohm  (center-tap) to 8 ohm transformer and it worked very very nicely - maybe better than the 2k-to-10k XFMR.

Using the pre-amp from the R1/R2 and a 301 Op-amp amplifier from my venerable "Solid State Basics for the Radio Amateur"....and, amazingly enough, got more gain than I need.  Gonna swap to the pre-amp that W7EL used in his classic "Optimized QRP Transceiver"

I have pre-built audio filter boards...W7ZOI's "Peaked Low-Pass filter, and a 600Hz bandpass filter from ARRL handbook formulas.  Will introduce them into the circuit when I get the basic gain blocks established.

Better than Legos!

Jerry
KI4IO
Warrenton, VA
13196 2016-10-03 18:05:39 davidpnewkirk Re: Gilbert Cell output
Jerry wrote:

Thanks all for your advice regarding my build of a discrete component NE602 (Gilbert Cell) as shown in EMRFD (figure 5.28)

[and more, deleted]

That's good stuff. I haven't yet built up the gumption to do that full discrete Gilbert cell, but another discretes-intead-of-an-IC EMRFD circuit _is_ alive and kicking at W9BRD: "A Direct Conversion Receiver Using a Discrete Component Product Detector" (EMRFD section 1.98, Figures 1.18 and 1.19). Its core is essentially a discrete-components equivalent to the CA3028/3053 differential-amplifier product detector that served as many hams' introduction to direct conversion in the late 1960s and early 1970s.

My version shares a pencil tin and two-gang FAST and SLOW tuning capacitors with a triode-connected-6AK6s-based regenerative receiver ( pictures -- minus added SLOW tuning cap -- at http://dpnwritings.nfshost.com/ej/pictures/pictures3.htm ; circuit of the regen at http://dpnwritings.nfshost.com/ej/wr/ ).

As Jerry is discovering of his discretes-based Gilbert cell, these simple receivers aren't just tune-around-and-hear-stuff toys; they can be quite sufficient for two-way communication if you give them a chance with decent peaked or low-pass audio filtering. _Sure_ they exhibit limited AM rejection and intermod resistance, but those effects can be significantly reduced by resorting to the MF/HF receiver user's most fundamental receiver-improvement tools: RF attenuation and input band-limiting, both of which aim to reduce ill effects moderately to greatly while compromising reception little or not at all.

A statement by Rick, KK7B, on page 9.36 in EMRFD's "Phasing Receivers and Transmitters" provides a fundamental clue to the potential value of input attenuation: "The direct conversion receivers described by the author in _QST_ in 1992-1995 were all developed using a full-sized elevated 40-m dipole in a quiet lakeside location in the Upper Peninsula of Michigan. At this location, signals from all over the US and Canada were quite strong, and the antenna noise power was always high enough that a 15-dB noise figure was always adequate."

Full stop.

"A 15-dB noise figure was always adequate." The passage goes on to more fully qualify necessary noise figure for location and frequency, but there's pivotal New You Can Use for just this snippet if at your band(s) of interest, disconnecting your receiver's antenna makes your receive "go dead" or "go quiet," and that news reads like this: If disconnecting your antenna makes your receiver go significantly quieter, _you are squandering the weaker-signal end of your receiver's dynamic range, including AM rejection, because your receiver is not only more sensitive than it need be, but more sensitive than it _should_ be.

The simpler, the more modest, the receiver, the more important this point. If your antenna noise floor comes in 15 or 20 dB above your receiver noise floor, run, don't walk, to adding sufficient RF attenuation between your receiver and antenna system such that disconnection the antenna makes it go distinctly, but not greatly, quieter. You will have push 3rd-order IMD down (assuming a system "well behaved" in terms of IMD) by a factor of 3x-the-added-attenuation; 2nd-order IMD, 2x. AM rejection--harder to easily quantity and commonly subject to relatively sharp threshold effects--will also be significantly improved.

It turns out, for instance, that for most of their history regenerative detectors have been applied pretty badly by those who depended on them. If you build your modern regenerative detector with a low-Z tank--100 ohms or less of X of L and C at the operating frequency--_and_ you adjust its signal diet such that antenna-system noise floor is just above the detector noise floor, you will find, as I do, that even at low beat notes (I like 300 to 600 Hz, preferably below 500 Hz) your detector won't frequency-pull before the loudness of received-signal audo makes you want to turn down your AF gain or (preferably) further reduce RF input.

With the four-6AK6s regenerative receiver linked above, 18 dB of so of attenuation is right for my location; with its 2N3904s-based-differential-amplifier pencil-tin-mate, somewhat less (and for it I also built in a DAY/NIGHT switch because my antenna noise floor differs enough between daytime and nighttime to warrant it). Again, however, this is at 40 meters in my particular suburbanish northern New Jersey location.

Paradoxically, simple receivers are even more fun when we take them seriously.

Best regards,

Dave
amateur radio W9BRD







 
13197 2016-10-05 14:04:09 jwolczanski Re: Gilbert Cell output
".....Paradoxically, simple receivers are even more fun when we take them seriously......".  Amen.

Funny, I was just going back over old articles and tumbled across "Direct Conversation - A neglected Technique" by Hayward and Bingham from the November 1968 issue of "QST".  That article could be re-printed today and nobody would guess it's almost 50 years old!

My receiver is done, just moving two circuits from perf board to PCB -  changed to the audio amplifier in EMRFD (figure 9.74).  Quite an impressive circuit, headphone rattling high-fidelity!

So Happy Birthday Direct Conversion.  I know you're older than fifty!

Jerry
KI4IO
13199 2016-10-05 18:30:02 davidpnewkirk Re: Gilbert Cell output
Jerry wrote:

Funny, I was just going back over old articles and tumbled across "Direct Conversation - A neglected Technique" by Hayward and Bingham from the November 1968 issue of "QST".  That article could be re-printed today and nobody would guess it's almost 50 years old!

_And_ that receiver is a fun first/early project, being based as it is entirely on discrete devices that one way or another are readily available today. (Yes, I built one, too, in the early 1970s.)

D-c is about a _century_ old. Harold H. Beverage in the late 19teens or early 1920s used a tikker, in series with headphones, with the pair connected between a very, very long wire and ground, to receive on Long Island VLF and LF CW signals from Europe. An LO-switched diode mixer is just a tikker that doesn't wear mechanically; instead of a tikker's RPM, we adjust the frequency of the oscillator that switches the diodes on and off.

Another form of direct conversion was the use of a low-power arc--100 W or so!--in close proximity to one's rectification-detection receiver (in effect, a "crystal" set, or, a bit later, a vacuum tube or sodium-vapor tube operating as a rectifier). The arc served as a powerful BFO.

And then Armstrong developed the regenerative detector, an implementation (when oscillating) of direct conversion self-generates its LO signal that combines considerable amplification at RF (by means of the oscillatory feedback) with modest amplification of the AF resultant.

Before hams generally went over the the superhet--which for CW reception puts a bunch of amplification and frequency conversion ahead of a d-c "second detector" usually consisting of little more than a thermionic diode and BFO--they were also using straight d-c in the form of the station _monitor_, a battery-powered oscillator (commonly a Type 30 triode) with headphones connected in series with its plate to hear the beat between the monitor's own signal and that of the station transmitter.

What we now so readily enjoy now as direct conversion--commonly based on a low-gain, even lossy, detector followed by high AF amplification--was long made practically impossible by the practical expedient of defaulting to heating one's tubes with a huge audio signal, which largelty disallowed today's high AF gains because of hum. Now it's a mainly dc world, and the RF and AF circuitry d-c entails can be small, cool, quiet, inexpensive, and efficient. D-c radios (er, and regens, sez I) just sound so _clean_...

Best regards, and congrats,

Dave
amateur radio W9BRD